Protein-protein interactions are essential to the normal physiological functions of cells and multicellular organisms. Many proteins in nature exhibit novel or optimal functions when complexed with one or more other protein chains. This is illustrated by various ligand-receptor combinations that contribute to regulation of cellular activity. Certain ligands, such as tumor necrosis factor .alpha. (TNF.alpha.), TNF.beta., or human chorionic gonadotropin (hCG), occur as multi-subunit complexes. Some of these complexes contain multiple copies of the same subunit. TNF.alpha. and TNF.beta. (collectively referred to hereafter as TNF) are homotrimers formed by three identical subunits (1-4). Other ligands are composed of non-identical subunits. For example, hCG is a heterodimer (5-7). Receptors may also occur or function as multi-chain complexes. For example, receptors for TNF transduce a signal after being aggregated to form dimers (8,9). Ligands to these receptors promote aggregation of two or three receptor chains, thereby affording a mechanism of receptor activation. For example, TNF-mediated aggregation activates TNF receptors (10-12).
The modulation of protein-protein interactions can be a useful mechanism for therapeutic intervention in various diseases and pathologies. Soluble binding proteins, that can interact with ligands, can potentially sequester the ligand away from the receptor, thereby reducing the activation of that particular receptor pathway. Alternatively, sequestration of the ligand may delay its elimination or degradation, thereby increasing its duration of effect, and perhaps its apparent activity in vivo. In the case of TNF, soluble TNF receptors have been primarily associated with inhibition of TNF activity (13-17).
Soluble binding proteins may be useful for treating human diseases. For example, soluble TNF receptors have been shown to have efficacy in animal models of arthritis (18,19).
Since TNF has three binding sites for its receptor (10-12), and dimerization of the cell surface receptor is sufficient for bioactivity (8,9), it is likely that binding of a single soluble receptor to TNF will leave open the possibility that this 1:3 complex of soluble receptor:TNF (trimer) can still bind and activate a pair of cell surface TNF receptors. To achieve an inhibitory effect, it would be expected that two of the receptor binding sites on the TNF trimer must be occupied or blocked by the soluble binding protein. Alternatively, the binding protein could block proper orientation of TNF at the cell surface.
Generally speaking, the need was felt of synthesizing proteins that contain two receptor (or ligands) chains, as dimeric hybrid protein. See Wallach et al., U.S. Pat. No. 5,478,925.
The primary strategy employed for generating dimeric or multimeric hybrid proteins, containing binding domains from extracellular receptors, has been to fuse these proteins to the constant regions of an antibody heavy chain.
This strategy led, for example, to the construction of CD4 immunoadhesins (20). These are hybrid molecules consisting of the first two (or all four) immunoglobulin-like domains of CD4 fused to the constant region of antibody heavy and light chains. This strategy for creating hybrid molecules was adapted to the receptors for TNF (10,16,21) and led to the generation of constructs with higher in vitro activity than the monomeric soluble binding proteins.
It is widely held that the higher in vitro potency of the dimeric fusion proteins should translate into higher in vivo activity. One study does support this, revealing an at least 50-fold higher activity for a p75(TBP2)-Ig fusion protein in protecting mice from the consequences of intravenous LPS injection (16).
However, despite the widespread utilization of immunoglobulin fusion proteins, this strategy has several drawbacks. One is that certain immunoglobulin Fc domains participate in effector functions of the immune system. These functions may be undesirable in a particular therapeutic setting (22).
A second limitation pertains to the special cases where it is desirable to produce heteromeric fusion proteins, for example soluble analogs of the heteromeric IL-6 or type I interferon receptors. Although there are numerous methods for producing bifunctional antibodies (e.g., by co-transfection or hybridoma fusions), the efficiency of synthesis is greatly compromised by the mixture of homodimers and heterodimers that typically results (23). Recently there have been several reports describing the use of leucine zipper motifs to guide assembly of heterodimers (24-26). This appears to be a promising approach for research purposes, but the non-native or intracellular sequences employed may not be suitable for chronic applications in the clinic due to antigenicity. The efficiency of assembly and stability post assembly may also be limitations.
On the other hand, in the particular case of TNF receptors, certain modifications to the p55 TNF receptor have been found to facilitate homodimerization and signaling in the absence of ligand (27,28). It has been found that a cytoplasmic region of the receptor, termed the "death domain," can act as a homodimerization motif (28,30). As an alternative to an immunoglobulin hybrid protein, fusion of the extracellular domain of the TNF receptor to its cytoplasmic death domain could conceivably result in a secreted protein which can dimerize in the absence of TNF. Such fusion proteins have been disclosed and claimed in the International Patent Application WO 95/31544.
A third further strategy employed for generating dimers of soluble TNF receptors has been to chemically cross-link the monomeric proteins with polyethylene glycol (31).